Download

The project is still on pre-release version, so there is no pre-packed version of the library. Still, you are welcome to visit the sourceforge page and browse the SVN repository, where the project is updated very regularly at the moment.

You can get the current source tree with the following command:
svn co https://eblearn.svn.sourceforge.net/svnroot/eblearn/trunk eblearn

Installation

Ubuntu

Required external packages (Non-Proprietary)

all packages: libeblearn, libidx, tool libraries, tester and demos
$ sudo apt-get install subversion cmake g++ libatlas-base-dev libboost-filesystem-dev imagemagick libboost-regex-dev libqt4-core libqt4-dev libqt4-gui libcppunit-dev
libeblearn and libidx only (tools, testers and some demos will not compile)
$ sudo apt-get install cmake g++ libatlas-base-dev imagemagick
for developers only (tools for bug-free development)
$ sudo apt-get install valgrind electric-fence

Compilation

$ ./configure
$ make
(or 'make debug' to compile with debug information for gdb)
(or 'make prj=idxgui' if you just want to compile the library libidxgui and its dependencies or 'make debug prj=tester' to compile the tester only in debug mode)

Execution

$ cd bin
$ ./tester
$ ./mnist "your-mnist-directory"

MAC OS X

(by George Williams from http://movement.nyu.edu/internal/Main/EblearnMacosNotes)
These are some notes that should help you get eblearn to compile and run on macos. They are based on revision 115 of the subversion source repository of the eblearn source.

Required external packages (Non-Proprietary)

Install cmake
o I installed from the src tarball (cmake-2.6.2.tar.gz) which you can get here at http://www.cmake.org/cmake/resources/software.html.
o Consult the README.txt in the top level directory and note that the steps to compile and install are "./bootstrap; make; sudo make install".
Install libatlas
o I installed from the src tarball (Atlas-C++-0.6.1.tar.gz) which you can get here at http://sourceforge.net/projects/worldforge/
o The standard "configure/make/sudo make install" will do.
Install cppunit
o I installed from the src tarball (cppunit-1.12.1.tar.gz) which you can get here at http://apps.sourceforge.net/mediawiki/cppunit/.
o The standard "configure/make/sudo make install" will do.
Install mnist databases
o You get these here http://yann.lecun.com/exdb/mnist/.
o You will need to unpack the data
o You will need the full path to the data directory later on
Install boost
o You should be able to install from the most recent boost version. I found a "macports" version, which is much easier to install.
o Macports install - If you don't have macports, use the tarball instructions here: http://www.macports.org/install.php. I installed from the tarball MacPorts-1.7.0.tar.gz.
o Macports boost - cd into the macports bin directory and type ">> sudo ./ports install boost". This may take a while.
Download and adjust eblearn build
o Download from the subversion repository: ">> svn export https://eblearn.svn.sourceforge.net/svnroot/eblearn/trunk eblearn"
o Modify the build to find boost headers. Locate the file "eblearn/scripts/FindBoost.cmake" and insert the line at the top of the file "SET(BOOST_INCLUDE_DIR /opt/local/include)". Assuming you did not change the default install directory of macports, then /opt/local/include will contain the macports versions of packages.

Compilation

o At this point, you can use the standard instructions to build and configure eblearn
o cd into the toplevel eblearn directory, and type ">>./clean.sh" (just in case you did a previous build with this src tree.)
o Then type ">> ./build.sh".
o Then type ">>./configure.sh". You'll need the full path to your local mnist data dir.

Execution

o cd into "bin" and type "./eblearn_tester".
o all the implemented unit tests should pass